cli-music_skill

Overview

This skill generates lo-fi synthesized electronic music entirely offline using CLI audio tools (ffmpeg and SoX). It produces short 1–3 minute tracks by building waveform layers (pads, bass, drums, texture) and mixing them into MP3 without any cloud services or API keys. Output sounds like chiptune / basic synth demos—no vocals or sampled instruments.

How this skill works

The skill synthesizes individual WAV layers using SoX waveforms or ffmpeg aevalsrc expressions (harmonic stacks, detune, vibrato, envelopes), creates noise-based percussion with anoisesrc, loops and sequences patterns, then mixes everything with ffmpeg filter graphs and exports a normalized MP3. It prefers SoX for native waveform generation and falls back to aevalsrc when SoX is not available.

When to use it

• You want to generate music locally with no network or API keys.
• You need quick lo-fi electronic or chiptune-style beds for prototypes or background use.
• You want a reproducible CLI workflow using ffmpeg and SoX.
• You need offline generation for demos, game jams, or embedded rigs.
• You want to learn expression-based synthesis (aevalsrc) and layering techniques.

Best practices

• Always stack harmonics (saw/square approximations) instead of a bare sine to avoid test tones.
• Keep frequency ranges separate: bass low, pads mid, melody high to avoid masking.
• Use per-note envelopes and short fades to prevent clicks when sequencing.
• Layer drum hits (tone + noise + shaping) rather than relying on a single source.
• Generate layers into a temp directory, mix with ffmpeg, then clean up temporary files.

Example use cases

• Create a 60s lo-fi instrumental loop for a game level using SoX pads and ffmpeg mixing.
• Prototype background tracks for a podcast segment without uploading audio anywhere.
• Produce chiptune-style beats for a demo reel using harmonic stacking and anoisesrc percussion.
• Batch-generate multiple variations by scripting BPM, scale, and chord progressions.
• Educate students on synthesis math by generating tones with aevalsrc expressions.